GMDGeoscientific Model DevelopmentGMDGeosci. Model Dev.1991-9603Copernicus PublicationsGöttingen, Germany10.5194/gmd-10-1665-2017Investigating soil moisture–climate interactions with prescribed soil moisture experiments:
an assessment with the Community Earth System Model (version 1.2)HauserMathiasmathias.hauser@env.ethz.chhttps://orcid.org/0000-0002-0057-4878OrthRenéhttps://orcid.org/0000-0002-9853-921XSeneviratneSonia I.Institute for Atmospheric and Climate Science, ETH Zurich, Zurich, SwitzerlandMathias Hauser (mathias.hauser@env.ethz.ch)20April2017104166516773August201627October20166March201719March2017This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://gmd.copernicus.org/articles/10/1665/2017/gmd-10-1665-2017.htmlThe full text article is available as a PDF file from https://gmd.copernicus.org/articles/10/1665/2017/gmd-10-1665-2017.pdf
Land surface hydrology is an important control of surface weather and
climate. A valuable technique to investigate this link is the prescription of
soil moisture in land surface models, which leads to a decoupling of the
atmosphere and land processes. Diverse approaches to prescribe soil moisture,
as well as different prescribed soil moisture conditions have been used in
previous studies. Here, we compare and assess four methodologies to prescribe
soil moisture and investigate the impact of two different estimates of the
climatological seasonal cycle used to prescribe soil moisture. Our analysis
shows that, though in appearance similar, the different approaches require
substantially different long-term moisture inputs and lead to different
temperature signals. The smallest influence on temperature and the water
balance is found when prescribing the median seasonal cycle of deep soil
liquid water, whereas the strongest signal is found when prescribing soil
liquid and soil ice using the mean seasonal cycle. These results indicate
that induced net water-balance perturbations in experiments investigating
soil moisture–climate coupling are important contributors to the climate
response, in addition to the intended impact of the decoupling. These results
help to guide the set-up of future experiments prescribing soil moisture, as
for instance planned within the Land Surface, Snow and Soil Moisture Model Intercomparison Project (LS3MIP).
Introduction
The interplay between the land surface and the atmosphere can induce or
modulate anomalies in temperature and
precipitation e.g.. Soil moisture (SM) is a key
quantity in this context . The complex role of SM in
land–atmosphere dynamics can be investigated with general circulation models
(GCMs). Typically in this context, land state variables are set – prescribed
– to predefined target values in GCM simulations. Such experiments have been
performed for decades e.g.. Prescribing land state
variables suppresses interactions between the land and the atmosphere and can
hence be used to infer the role of land–atmosphere interactions for the
climate.
The Global Land Atmosphere Coupling Experiment GLACE; was the first major multi-model effort to comprehensively analyse
the impact of SM on several atmospheric variables in the context of present
climate. In multi-model simulations of a particular Northern Hemisphere
summer, regions of coupling between precipitation and evaporation were
identified. While some regions emerged as multi-model “hot spots”, the
experiment revealed a large inter-model spread in the land–atmosphere
coupling strength, pinpointing the different sensitivities of the models with
respect to the link between SM and evapotranspiration, and the link between
evapotranspiration and precipitation .
More recently, the role of SM–climate feedbacks in climate change
projections has been investigated in the multi-model project Global
Land–Atmosphere Coupling Experiment of the Coupled Model Intercomparison
Project, phase 5 GLACE-CMIP5;. In GLACE-CMIP5, an
ensemble of GCMs performed two distinct experiments for the period 1950 to
2100 to assess the role of inter-annual SM variability, and of SM trends for
climate change simulations. The removal of both, interannual SM variability
and the long-term SM trend by prescribing the mean seasonal cycle from 1971
to 2000 “experiment A”;, leads to large
decreases in temperature extremes as well as effects on precipitation
extremes . In another
experiment (“experiment B”) the 30-year-running mean of the reference experiment is prescribed to preserve
long-term SM trends. Projected SM drying trends were found to be accompanied
by a further increase of temperature extremes. However, the simulated SM
trends were strongly model dependent.
In the context of the upcoming CMIP6 modelling cycle, the Land Surface, Snow
and Soil Moisture Model Intercomparison Project LS3MIP;
plans a variety of experiments to quantify and compare the role of multiple
land state variables in climate change simulations. Particularly, the Land
Feedback Model Intercomparison Project (LFMIP) within the LS3MIP
project plans experiments similar to the
GLACE-CMIP5 project, which aim to quantify the role of land–atmosphere
feedbacks at the climate timescale. In contrast to the GLACE-CMIP5
experiments, simulations will be run with an interactive ocean.
Additionally to the above-mentioned GLACE-type experiments, a large number of
studies analysed the influence of SM on the atmosphere from multiple
perspectives e.g.. The different goals, and
also the different employed land surface models in these studies motivated
and necessitated different techniques to prescribe SM. They include the
prescription of (1) all land state variables, (2) only SM at all soil depths,
(3) SM in subsurface soil layers only, (4) nudging SM values, and
(5) restricting the SM prescription to certain regions. In addition, the
prescribed SM values vary widely between studies. Some use the plant wilting
point and the field capacity to simulate extreme dry and wet conditions,
respectively. Others use simulated SM from a particular year, a
climatological seasonal cycle, or a smoothed seasonal cycle. Furthermore, the
SM climatology can be estimated (calculated) in different ways: either using
the meanas done in e.g. or the
medianas done in. A third
difference between the SM-prescription methodologies is the temporal
resolution of the SM target dataset – they comprise instantaneous, daily,
and interpolated monthly data.
Similarly to prescribing sea surface temperatures in GCMs, which does not
allow for conservation of the energy balance, modelling experiments
prescribing SM infringe the water balance of the land model. However, water
is only added or removed by the prescription algorithm within the soil and
not in the atmosphere or at the land–atmosphere interface. Thus, and because
such experiments analyse only the atmospheric response, the perturbation of
the soil water balance is “deemed acceptable” . Still,
prescribing SM induces artificial sources and sinks of water in the model. To
our knowledge a quantification of this water-balance disturbance and its
impact is currently lacking. In particular, the distinct effects of different
existing methodologies on these water imbalances and their impact have not
been systematically compared so far. This is an important gap because it is
possible that they could lead to methodologically induced discrepancies
between studies.
In the present article, we analyse differences in SM-prescribing set-ups that
aim to remove the inter-annual variability while conserving the seasonal
cycle of SM to assess its impact on surface climate. In this context, we
focus on methodologies that are relevant for the LS3MIP experiment such that
our conclusions can contribute to the final implementation of its
experimental design.
Model description
In this section we first introduce the employed GCM and the corresponding
land surface scheme. Thereafter, we describe the different tested approaches
to prescribe SM. Finally, we provide an overview of the conducted
experiments.
In this study, we use the Community Earth System Model CESM;version
1.2. This is a fully coupled Earth system model, combining
separate modules for the atmosphere, the ocean, and the land. Land surface
processes and their coupling to the atmosphere are simulated by the Community
Land Model, version 4.0 CLM4;. CLM4 is a
third-generation land surface model ,
incorporating the hydrological cycle (see below), land surface energy fluxes,
a variety of land surface types (wetlands, glacier, vegetated, etc.) and up
to 15 generic plant types (“plant functional types”), among others.
Short overview of hydrology in the Community Land Model
Water in CLM4 is stored in four reservoirs: on the canopy, as snow, as
groundwater, and in the soil. The soil is divided into 15 vertical layers
with exponentially increasing thickness from top to bottom. However, only the
10 first layers are hydrologically active and extend to a depth of 3.8 m
(the last five layers act only as thermal sink/source). Water reaching the
soil surface through precipitation and stemflow is partitioned into surface
runoff and infiltration, i.e. water entering the uppermost soil layer. Water
is removed from the soil by subsurface runoff (drainage) and canopy
transpiration through root extraction. The water flux within the soil is
governed by Darcy's Law. The corresponding hydraulic properties are a
function of soil water content and texture. Water can occur in liquid and
solid states, which will be referred to as LIQ and ICE for the remainder of
this study. A comprehensive description of CLM4 can be found in
.
Prescription of soil moisture in the Community Land Model
The aim of SM prescription is to control the soil's water content, i.e. to
force it to a predefined target value (e.g. a climatological seasonal cycle,
the plant wilting point or others), irrespective of the actual conditions in
the soil. As this is not possible with the default model version, we extend
the original model code of CLM4 with a module that
reads the target value from a previously prepared file and overwrites the
actual value in the model after each time step. The goal of this study is to
assess and compare various approaches of prescribing SM. The tested
techniques comprise established as well as novel methods as listed in
Table and Fig. .
The four tested approaches to prescribe SM in CLM. The target LIQ,
ICE, and SM values are denoted LIQt, ICEt, and
SMt, respectively. SMt corresponds to the sum of
LIQt and ICEt (i.e. SMt= LIQt+
ICEt). In general the target values depend on time (day of year),
location (grid point), and depth (soil level). In this study we use the
30-year-mean and 30-year-median seasonal cycle; however, other targets are possible,
e.g. a specific year. (a) LIQ and ICE are both prescribed in
PRES_LIQ+ICE. (b) In PRES_FRAC, total SM is prescribed, but the
fraction f=LIQ/(LIQ+ICE) is interactively
computed by the model. Note that the hydrology in CLM4 is still active.
(c) Illustration of the new approach (PRES_LIQ), prescribing LIQ in
all soil levels if the soil temperature is above freezing (left) and for an
example with soil level two below freezing (right).
(d) PRES_LIQ_DEEP as PRES_LIQ, but the first soil layer is always
interactive.
In previous studies SM in CLM
was prescribed by setting LIQ and ICE individually to the predefined values
at each time step (Fig. a). This technique will be referred to
as PRES_LIQ+ICE. A second technique, named PRES_FRAC
(Fig. b) also prescribes LIQ and ICE, but lets the land surface
model interactively compute the fraction of LIQ e.g. applied
in. Hence, the model has an additional degree of freedom
compared to PRES_LIQ+ICE.
Furthermore, we propose an alternative approach where SM is only prescribed
when the soil temperature is above 0∘C (PRES_LIQ). If the soil
is frozen, LIQ and ICE are both computed interactively. The climatological
total SM (i.e. LIQ + ICE) is converted into LIQ for the prescription. The
important characteristic of this new algorithm is that it never artificially
adds ICE (see Sect. ). Although (supercooled) LIQ and ICE can
coexist in CLM4, we leave the soil hydrology entirely interactive below the
freezing temperature. In detail the algorithm works as follows: LIQ is
prescribed starting from the uppermost soil level, and then further down
until either the soil bottom is reached, or until a layer with soil
temperature at or below 0∘C is found (Fig. c). This
follows the methodology employed in the (optional) irrigation module of CLM
.
Following an approach presented in , and also used in
, we additionally test a similar methodology as in
PRES_LIQ, but without prescribing the topmost soil layer, hereafter named
PRES_LIQ_DEEP (Fig. d). Whereas in the other prescription
approach the land–atmosphere coupling is entirely removed, this allows for a
limited feedback between the soil and the atmosphere. Even though the topmost
layer is only 1.8cm thick, it controls bare-soil evaporation,
which forms a significant part of the total evapotranspiration. Additionally,
SM in the topmost layer – in contrast to the deep(er) soil layers – may not
be well predictable as it does not have its considerable inertia and memory
.
For all four methods the hydrology in CLM4 is still active – SM is removed
by root extraction and drainage and added by infiltration. However, at the
end of each time step, this interactively calculated SM is overwritten and
set to the target value. We record the difference of the interactively
computed SM and its target value as the water-balance perturbation. If it is
positive, the algorithm has artificially “added SM”, while it has “removed
SM” if the difference is negative.
Finally, we have to choose the time resolution of the SM dataset from at
least four possibilities: (1) monthly data with linear interpolation to daily
mean values, (2) daily mean values, (3) daily mean values with linear
interpolation to every model time step, and (4) instantaneous values at every
model time step. In this study we use daily mean values, as
linearly interpolated monthly values can be too coarse (see below).
Overview of the experiments
All simulations (Table ) are conducted with CESM. As reference
simulation we perform a fully coupled simulation from 1950 to 2099 (hereafter
called REF), combining the historical forcing and the Representative
Concentration Pathway 8.5 scenario RCP8.5;. The
daily SM output from REF between 1971 and 2000 is used to calculate the mean
and median climatology at every grid point, soil level, and day of the year
for LIQ and ICE individually.
We perform seven simulations with prescribed SM that differ in the method to
prescribe SM (Sect. and Fig. ), and the
target SM climatology. In the simulations with prescribed SM, we also
prescribe sea surface temperatures (SSTs) and sea ice from REF to suppress
impacts from changed SSTs in response to the prescribed SM as done in
GLACE-CMIP5;. The first two simulations (PRES_LIQ_MEAN
and PRES_LIQ_MEDIAN) use the new SM-prescription scheme described above
with mean and median climatologies, respectively. The third simulation,
PRES_LIQ_DEEP_MEDIAN, also uses the new prescription scheme but leaves the
first layer interactive. In the fourth and fifth simulations
(PRES_LIQ+ICE_MEAN and PRES_LIQ+ICE_MEDIAN), we prescribe LIQ and ICE and
also compare mean and median SM climatology. Finally, simulations six and
seven also prescribe LIQ and ICE, but calculate the respective fractions
interactively (PRES_FRAC_MEAN and PRES_FRAC_MEDIAN). In our analysis we
concentrate on the simulations that do not prescribe ICE because
both techniques that do so lead to large, unrealistic surface temperature and
ground heat flux anomalies (see
Sect. ). The variety of soil moisture prescription approaches
considered here makes this study a valuable basis for the final planning of
the soil moisture prescription methodology for the LS3MIP simulations.
Results and discussionSoil moisture climatologyMean vs. median soil moisture
The daily mean and median SM climatologies only differ if the inter-annual SM
values are not symmetrically distributed. As an example, Fig. a
shows the evolution of SM throughout the year in the topmost 10cm
of the soil for a location in India. This grid point shows a distinct
seasonal cycle with a dry period from February to May and high soil-moisture
values during the rest of the year. In the dry season the median is generally
smaller than the mean, with large rainfall events leading to outliers on the
wet end of the distribution. For example on the 5 April, the difference is
-2.3mm, or -14.0% (Fig. b). During the wet
period the median is usually larger than the mean, and it is dry years that
lead to the asymmetry. However, the difference between median and mean are
generally smaller compared to the dry
period; e.g. on the 21 December it is 1.0mm, or 3.8%
(Fig. c). There are many processes that contribute to
non-symmetric SM distributions: the positively skewed distribution of
precipitation, the upper and lower bound in the water holding capacity of the
soil (between the wilting point and saturation), as well as the strong
non-linear function of water flow (hydraulic conductivity) within the soil
with respect to the SM state .
Difference between mean and median SM. (a) Seasonal cycle
of total SM in the top 10cm for an example grid point in India
(10.4∘ N, 77.5∘ E) as simulated by CLM for the
climatological period (1971 to 2000). Shown are the individual years (grey
lines), and their mean (red) and median (blue). Light grey background shows
the 3 consecutive hottest months at this grid point and vertical black
lines the 2 days depicted in (b) and (c), respectively.
(b) and (c) Kernel density estimate of the SM distribution
(thick black line), including the individual years (thin grey lines) and the
mean (red) and median (blue) SM values for the 5 April (b) and
21 December (c). (d) and (e) Relative difference
in the SM climatology between median and mean for the hottest months of the
year in the surface layer (0 to 10cm, d) and in 10 to
100cm depth (e).
In Fig. d and e, we show the relative difference between the
mean and median SM for two depth intervals. We thereby focus on the 3 hottest
consecutive months of the year, as we expect SM differences in these months
to have the largest temperature impact. The hottest months of the year are
determined from REF. On land in the mid- and high latitudes these 3 hottest
consecutive months generally correspond to summer, i.e. June to August in
the Northern Hemisphere and December to February in the Southern Hemisphere
(Fig. S1 in the Supplement). The largest relative differences between the
mean and median are found in the uppermost 10cm of the soil
(Fig. d). Regions for which the median is drier than the mean
include Australia, northern Africa, the Mediterranean, and western North
America, while it is wetter in
central Africa, central Europe, western Asia and central North America. As
for the example grid point in India, negative differences are generally
stronger than positive differences. In contrast to these large relative SM
differences in the top 10cm, the relative differences are
generally below 2% in depths between 10 and 100cm
(Fig. e), and from 100 to 380cm (not shown). The
absolute differences, however, are higher for deeper soil levels, as these
are thicker. A difference between the mean and median climatologies in the
topmost 10cm of the soil is not only a feature of CESM but also
evident in other models participating in GLACE-CMIP5 (Fig. S2).
Daily vs. interpolated monthly soil moisture
In this study we prescribe daily SM values, whereas some previous studies used
daily values obtained from a linear interpolation of monthly means
e.g. some simulations in the GLACE-CMIP5
experiment;. True daily and interpolated monthly SM
values can differ in regions with a short sharp peak in the seasonal cycle,
as exemplified for a grid point in central Africa (Fig. a). It
shows true daily values (blue line) and the corresponding monthly means (blue
dots). The orange line illustrates daily values linearly interpolated from
the monthly mean values, where these monthly values were assumed to occur in
the middle of each month. While true daily and interpolated monthly values
match closely for most of the year, the latter does not entirely capture the
summer minimum. In addition, the monthly means derived from the interpolation
(orange dots) are not equal to the true monthly means derived from the daily
time series. In contrast, the annual mean of the daily and monthly
interpolated values are equal.
Difference between interpolated monthly and daily SM.
(a) Seasonal cycle of median SM climatology for one grid point in
central Africa (0.9∘ N, 25∘ E), illustrating the difference
between daily and interpolated monthly values. (b and
c) Absolute difference
[%] in the median SM climatology between daily and interpolated
monthly values in the surface layer (0 to 10cm, b) and
in 10 to 100cm depth (c).
We show the median absolute differences of the warm season months between
true daily and interpolated SM values in Fig. b and c. While the
difference is generally smaller than between mean and median SM
climatologies, it is comparable in some regions, e.g. the Sahel, southern
Africa, and Australia (Fig. b). For the depth intervals 10 to
100cm, and 100 to 380cm (not shown), the relative
difference is generally below 2%. In contrast to the difference
between the mean and median SM climatologies, positive and negative
deviations between daily and interpolated monthly SM climatologies compensate
when integrated over time. This analysis shows that other methodological
differences apart from using mean or median seasonal cycle may (regionally)
cause important implications.
Temperature responsePrescribing soil liquid water only
Difference in the median of the simulation with prescribed SM and
REF (anomaly) for the period 1971 to 2000. Panels (a) to (c)
annual-mean temperature, panels (d) to (f) TXx. Significance is
tested with a Wilcoxon–Mann–Whitney U test e.g..
Conducting a significance tests at each grid point increases the probability
to falsely reject the null hypothesis e.g.. We therefore
control this with the approach described by , using
a global p value of 5%.
In this section we investigate the influence of the newly developed SM-prescription
methodologies on surface air temperature. Figure a
to c show the climatological temperature between 1971 and 2000 for each
methodology compared to REF. In all three simulations the mean land
temperature is lower than in REF. The largest difference is found for
PRES_LIQ_MEAN, which has negative temperature anomalies for almost all land
grid points. PRES_LIQ_MEDIAN has smaller temperature anomalies than
PRES_LIQ_MEAN, corresponding to the regions with smaller climatological SM
when comparing the median to the mean (Fig. ). For
PRES_LIQ_DEEP_MEDIAN we obtain the smallest anomalies. We find similar
results when comparing the experiments to REF for the time period 2070 to
2099 (Fig. S3a to c). Thus, the global land warming between 1971 to 2000 and
2070 to 2099 is only slightly larger in REF than the experiments. This is in
line with earlier findings , although experiments in
this study are at the lower end of the range of the individual GLACE-CMIP5
models.
In addition to changes in annual-mean temperature in response to prescribed
SM, we also investigate corresponding changes in annual maximum daily maximum
temperature (TXx), shown in Fig. d to f. In most regions the TXx
differences are larger than the annual-mean differences. This stronger impact
of SM changes on extremes vs. mean temperatures is a well-known
characteristic of land–atmosphere coupling e.g.. TXx in PRES_LIQ_MEDIAN are cooled by more than
2∘C by the SM prescription for Australia, southern Africa,
India and Brazil. The results for PRES_LIQ_DEEP_MEDIAN are similar to
PRES_LIQ_MEDIAN, except in southern Australia and northern high latitudes.
These results are in line with earlier studies e.g..
The cooling increases towards the end of the 21st century in all three
simulations (Fig. S3d to f).
Prescribing soil ice
In this section we analyse PRES_LIQ+ICE_MEAN and PRES_LIQ+ICE_MEDIAN,
i.e. the simulations that prescribe ICE. Using the PRES_LIQ+ICE methodology
leads to a similar anomaly in global-land mean temperature in the 1971 to
2000 period than prescribing LIQ only (PRES_LIQ+ICE_MEAN:
-0.8∘C and PRES_LIQ+ICE_MEDIAN:
-0.3∘C, Fig. S4). However, these land temperature
differences increase strongly toward the end of the 21st century
(Fig. S5), in contrast to the
simulations without prescribed ICE. As the climate and hence the soils warm,
the soil ice melts, and, as the ICE climatology is based on the time period
1971 to 2000, more soil ice is prescribed. Consequently, melting occurs
during every modelling time step and the soil ice is re-prescribed at the end
of the time step, thereby constantly cooling the land surface and hence
near-surface temperature. Thus, prescribing soil ice leads to a strong
disturbance of the model's energy balance. This is also evident in the large
ground heat flux anomalies of the
simulations with prescribed ICE (more than 10Wm-2 locally
and 1.9Wm-2 globally for 2070 to 2099, Fig. S6). In
contrast, experiments that do not prescribe ICE do not show any noteworthy
ground heat flux anomalies. As the
climate warms, there is an increasing land area where the air temperature is
no longer consistent with a frozen ground; thus, the land mean temperature
anomaly increases with time. The largest temperature signal occurs locally in
the mid- and high latitudes. However, non-local effects due to heat advection
and/or altered atmospheric circulation can not be excluded.
Note that most climate models, for instance within GLACE-CMIP5, do not
prescribe ICE and thus do not suffer from this problem. However, ICE was
prescribed in CESM in earlier studies . This may have caused an increased temperature perturbation
that does not affect the main conclusions of these studies. In the GLACE
experiments, simulated a summer in the current climate,
which reduces the influence of prescribing ICE. Additionally, they
concentrated their analysis on the variability of precipitation on non-ice
land points. compared two simulations that both
prescribe ICE, such that the effects cancel while others excluded CESM
simulations from their analysis e.g..
Interactive fraction of liquid and frozen soil water
The last two simulations, PRES_FRAC_MEAN and PRES_FRAC_MEDIAN, prescribe
total SM while the relative proportions of LIQ and ICE are interactively
computed by the model. Hence, this technique should circumvent the problem of
repeatedly adding and melting ICE. However, due to vertical liquid water
transport in the soil it also leads to large temperature and ground
heat flux anomalies in CLM4
(Figs. S4 and S6). In contrast to PRES_LIQ+ICE the annual-mean temperature
anomaly is already apparent for the period 1971 to 2000 and increases only
slightly toward the end of the 21st century (Fig. S5). Nonetheless, we think
that this technique is viable, and that the problem reported here is CLM4
specific. For example, Fig. 2 in gives no indication of
a large temperature anomaly due to the prescription of ICE. It is recommended
to calculate the ground heat flux
anomalies when prescribing SM, as this is a good indicator of ICE-induced
energy balance perturbations.
Amount of prescribed soil moisture
SM is usually prescribed to suppress the land–atmosphere
coupling. This comes at the cost of water-balance perturbations. To quantify
the introduced imbalance, we separately compute the total of
(intentionally) added and removed SM for
all simulations with respect to REF (for which it is zero). During 1971 to
2000, the average amount of added SM (over the whole soil column) is about
650mmyr-1 (not shown). This is about three-quarters
of the global-land mean precipitation in REF. However, a similar amount of SM
is removed and the net water-balance perturbation is much smaller because
positive and negative perturbations largely compensate when integrated over
the entire soil column. A large amount of water is usually removed from the
uppermost soil layers because rain infiltrates the topmost soil layer but has
not enough time to reach deeper soil layers before this wet SM is replaced
with a (usually) drier climatological value at the end of the time step.
Consequently, the deeper layers are too dry and water is added by prescribing
the climatological SM.
Mean annual SM perturbation for 1971 to 2000. Panels (a) to
(c) net water-balance perturbation in mmyear-1,
panels (d) to (f) net water-balance perturbation scaled by the
annual-mean precipitation.
For these reasons we focus on the net water-balance perturbations in the
remainder of this section. In PRES_LIQ_MEAN (Fig. a),
comparatively large amounts of water (>250mmyr-1)
are added in Australia, India, mainland Southeast Asia (Indochina), southern
Brazil, and parts of Africa. The regions with large amounts of net added SM
coincide with regions where we find the strongest TXx reductions in
Fig. , a consequence of the (muted) land–atmosphere coupling.
These regions show large positive anomalies in evapotranspiration, which is
responsible for the large amounts of added LIQ, as well as the reduction of
the sensible heat flux, which in turn leads to lower TXx. Interestingly, TXx
decreases at almost all land grid
points, while in many regions more water is removed than added. This is
explained by evapotranspiration, which increases in most land areas (not
shown), thus indicating that the SM prescription ensures availability of
water even during hot and dry periods. To set the water-balance perturbations
into perspective, we scaled the amount of net SM changes by the annual-mean
precipitation at each grid cell (Fig. d for PRES_LIQ_MEAN). In
many regions, the net water-balance perturbation is more than 30%
of the annual-mean precipitation amount (Fig. d). Not
surprisingly, we find the largest relative changes in regions with large
absolute SM changes, but also regions with small precipitation amounts
(Sahara, Arabian Peninsula).
Simulations with prescribed median SM generally display smaller water-balance
perturbations. In PRES_LIQ_MEDIAN (Fig. b and e), the net
water-balance perturbation is generally below
200mmyr-1. This corresponds to a perturbation of
less than 15% of annual-mean precipitation in most regions.
Regions where less water is added in PRES_LIQ_MEDIAN than PRES_LIQ_MEAN
also show substantially smaller evapotranspiration, because the median SM
climatology is smaller than the mean. On the other hand, regions where more
water is added with the median SM climatology, often show more rainfall,
especially northern Brazil.
Results for PRES_LIQ_DEEP_MEDIAN (Fig. c and f) are similar,
with the exception that the land area where SM amounts larger than
30% of annual-mean precipitation are removed is strongly reduced,
probably because water infiltrated in the topmost layer is evaporated (or
persists in this layer) instead of removing it with the algorithm.
Time series of global-land, annual-mean (a) net-prescribed
soil moisture, (b) precipitation, (c) evapotranspiration,
and (d) total soil moisture content. Total soil moisture in the
simulations with prescribed soil moisture is not entirely constant because
ICE is still computed interactively. The light grey background shows the two
time periods used for the climatology.
In terms of global net SM changes, PRES_LIQ_DEEP_MEDIAN introduces the
smallest water-balance perturbation of all simulations,
(-2mmyr-1, during 1971 to 2000). This is only
slightly more in the case of PRES_LIQ_MEDIAN
(-5mmyr-1). We find stronger water-balance
perturbations in PRES_LIQ_MEAN (43mmyr-1). Note
that in individual years, the water-balance perturbations can be larger
(Fig. a). Until the middle of the 21st century these
perturbations are relatively constant for all three simulations and decrease
thereafter. Thus, the small negative anomalies in PRES_LIQ_MEDIAN and
PRES_LIQ_DEEP_MEDIAN become about -45mmyr-1 for
2070 to 2099. For PRES_LIQ_MEAN, on the other hand, the large positive
water-balance perturbations decrease to 5mmyr-1.
This is caused by increased rainfall over land, which is only partially
compensated by increased evapotranspiration (Fig. b and c). The mean SM climatology is generally
wetter than the median climatology, this brings the
interactively computed SM closer to the mean climatology, such that less
water-balance perturbations are introduced by the SM prescription.
Consequently, there is also an increase in global-land mean total SM in REF
(Fig. d) in the CESM model. Note that this stands in contrast to
other models , which mostly display drying trends over land.
In these models, the water-balance perturbation for prescribing the mean SM
climatology would probably increase and not decrease in the future. Thus, on
global maps of net water-balance perturbations for 2071 to 2100 (Fig. S7),
the regions with large amounts of added SM are similar to those shown in
Fig. a, but more regions show larger amounts of removed SM.
Summary of the findings and recommendations for prescribing soil
moisture in land surface models.
Whole column vs. subsurface prescription of soil moisturePrescribing soil moisture in subsurface soil levels only, rather than the entire soil column, leads to a marginally smaller water-balance perturbation and atmospheric response.Soil moisture climatology (median vs. mean)Prescribing the median rather than the mean soil moisture leads to a considerably smaller perturbation of the water balance and also of the atmospheric response.Temporal resolution of the soil moisture climatology (daily vs. monthly soil moisture values)Daily soil moisture follows the seasonal cycle more closely and avoids the difference in monthly means of the reference simulation and the simulation with prescribed soil moisture. While not tested with simulations in this study, the differences in terms of water-balance and temperature perturbations when prescribing true daily vs. interpolated monthly SM (see Sect. ) may regionally be as large as the ones we find between prescribing mean vs. median seasonal SM cycles.Water-balance perturbation as outputWe recommend to output the amount of water that is added/removed by the algorithm as this may help to disentangle the water-balance perturbation and the land–atmosphere coupling.Prescribing soil icePrescribing soil ice leads to large temperature and ground heat flux anomalies. To prevent such anomalies the soil moisture prescription should be stopped as soon as the soil reaches freezing temperature. It should thus be ensured that the ice (or water to ice ratio) in the soil can evolve freely. If soil ice should nevertheless be prescribed, using a running median of soil ice and liquid of the control simulation will lead to the smallest perturbations.Conclusions
Soil moisture is commonly prescribed in general circulation models to study
the interplay of the land surface with weather and climate. As other types of
sensitivity experiments (e.g. prescribing sea surface temperatures), this
approach introduces perturbations, in particular to the land water balance,
because it artificially removes rainwater that infiltrates the soil and
replaces water in the soil that is lost via evapotranspiration and drainage.
It is important to be aware of these perturbations because they induce
changes in the surface climate and constitute a substantial fraction of the
climate response to the prescribed soil moisture conditions. Thus,
independent experiments investigating the impact of soil moisture–climate
interactions may come to different conclusions if they use different
approaches to decouple the land surface. However, perturbing the water
balance is necessary and cannot be avoided when aiming at an estimation of
the land–atmosphere coupling strength. Therefore, we investigate the impact
of different prescription techniques on climate and, for the first time,
also report the water-balance perturbations induced by soil moisture
prescription.
We implement and test four approaches to prescribe soil moisture, and use two
methods to estimate the soil moisture climatology (mean and median) in the
Community Earth System Model (CESM) with its land component, the Community
Land Surface Model (CLM). We show that the mean and median soil moisture
climatologies differ, with the most notable relative differences in the
uppermost soil layers. This difference is also observed in other general
circulation models within GLACE-CMIP5.
The first method to prescribe soil moisture that was originally developed for
CESM/CLM prescribes not only soil liquid water but also soil ice e.g. simulations contributing to GLACE experiments;. This leads to
large anomalies in the ground heat flux and the global mean temperature,
especially toward the end of the 21st century, and is therefore generally not
recommended. Similar problems are apparent in CLM (version 4) when total soil
moisture is prescribed while computing the relative proportions of soil
liquid water and soil ice by the model. We propose an alternative methodology
where no soil moisture is prescribed if the soil temperature in a particular
layer is below freezing point, and only soil liquid water is prescribed
otherwise. This method remedies the large global mean temperature and ground
heat-flux bias of the first method, while it still allows one to mute the
land–atmosphere coupling. For this method, we compare the difference between
using the mean and the median soil moisture climatology. When prescribing the
mean climatology, large net water-balance perturbations arise (global-land
mean of 50mmyr-1, for 1971 to 2000). Whereas in the
case of prescribing the median soil moisture climatology, the land mean
water-balance perturbation is much smaller
(-5mmyr-1). Thus, prescribing the median soil
moisture climatology leads to a considerably smaller perturbation of the
water balance. However, long-term soil moisture trends may also influence the
water-balance perturbations when prescribing a fixed (past) SM climatology.
This illustrates the utility of reporting the water-balance perturbations,
which is also planned within the LS3MIP project .
Corresponding to different water-balance perturbations, there are different
impacts on temperature; when prescribing the mean soil moisture climatology
we find a land mean cooling of more than 0.5∘C, while
prescribing the median leads to a mean land cooling of only
0.3∘C. Regionally, temperature differences of
2∘C are observed when prescribing the two climatologies.
Our results allow one to disentangle the influence of the soil
moisture–temperature coupling and the influence of the water-balance
perturbation.
For comparison, we furthermore test another well-established method
to prescribe soil moisture where the topmost
soil layer is computed interactively and soil moisture is only prescribed in
the lower layers. Results with this method are very similar to the findings
obtained when prescribing the whole soil column. Due to the interactive top
layer, the water-balance perturbation, and also the temperature signal are
slightly smaller.
This study shows that a careful design of the soil moisture prescription
methodology can help to minimize its influence on the model climate.
Therefore, Table provides a summary of our findings, and
recommendations for the set-up of studies prescribing soil moisture. These
recommendations can guide the implementation of the LFMIP experiments within
the LS3MIP project. Particularly, the method to prescribe soil moisture is not specified
within the LS3MIP project and this paper can serve as reference for model developers. We
note that the originally planned LS3MIP set-up mentions the use of the mean
climatological soil moisture . Using the median climatology
as recommended in this study would require a small adaptation of the
protocol, but this may still be possible as the simulations have not yet
started.
As the land–atmosphere coupling is removed in all experiments in this study,
the observed differences in the temperature signals are solely related to
differences between the induced water-balance perturbations. While these
perturbations are inevitable for suppressing the land–atmosphere coupling,
our results suggest that the role of these perturbations for the resulting
temperature signal is not negligible. Hence, not the entire temperature
signal can be attributed to the land–atmosphere coupling. This problem can be
addressed by prescribing the median SM climatology, which helps to reduce
water-balance perturbations because of the non-symmetrically distributed SM
in many regions.
The used code is available at
https://github.com/IACETH/prescribeSM_cesm_1.2.x, where the
documentation is linked. The code is released under a MIT licence. Revision
67cf64 was used to conduct simulations 1 to 5 and revision c38753 for
simulations 6 and 7. Note that the model framework (and code) of CESM/CLM is
necessary to compile and use the code given in the repository.
The Supplement related to this article is available online at doi:10.5194/gmd-10-1665-2017-supplement.
M. Hauser mainly performed the analysis and wrote the paper.
All authors participated in the design of the experiments, discussion of the results and writing of the paper.
The authors declare that they have no conflict of
interest.
Acknowledgements
This research was funded by the ERC DROUGHT-HEAT project (contract no.
617518). We thank Ruth Lorenz for discussion of the paper and Urs
Beyerle for support with CESM. Parts of the employed source code was
originally developed by Ruth Lorenz and Dave
Lawrence.Edited by: D. Roche
Reviewed by: B. van den Hurk and J. Colin
References
Benjamini, Y. and Hochberg, Y.: Controlling The False Discovery Rate – A
Practical And Powerful Approach To Multiple Testing, J. Roy.
Stat. Soc. B Met., 57, 289–300, 1995.Berg, A., Findell, K., Lintner, B., Giannini, A., Seneviratne, S. I., van den
Hurk, B., Lorenz, R., Pitman, A., Hagemann, S., Meier, A., Cheruy, F., Durcharne, A., Malyshev, S., and Milly, P. C. D.:
Land-atmosphere feedbacks amplify aridity increase over land under global
warming, Nature Climate Change, 6, 869–874, 10.1038/nclimate3029,
2016.Conil, S., Douville, H., and Tyteca, S.: The relative influence of soil
moisture and SST in climate predictability explored within ensembles of AMIP
type experiments, Clim. Dynam., 28, 125–145,
10.1007/s00382-006-0172-2, 2007.Douville, H.: Assessing the influence of soil moisture on seasonal climate
variability with AGCMs, J. Hydrometeorol., 4, 1044–1066,
10.1175/1525-7541(2003)004<1044:ATIOSM>2.0.CO;2, 2003.Douville, H., Chauvin, F., and Broqua, H.: Influence of soil moisture on the
Asian and African monsoons. Part I: Mean monsoon and daily precipitation,
J. Climate, 14, 2381–2403,
10.1175/1520-0442(2001)014<2381:IOSMOT>2.0.CO;2, 2001.Douville, H., Colin, J., Krug, E., Cattiaux, J., and Thao, S.: Midlatitude
daily summer temperatures reshaped by soil moisture under climate change,
Geophys. Res. Lett., 43, 812–818,
10.1002/2015GL066222, 2016.Fischer, E. M., Seneviratne, S. I., Luethi, D., and Schaer, C.: Contribution
of land-atmosphere coupling to recent European summer heat waves,
Geophys. Res. Lett., 34, 6, 10.1029/2006GL029068,
2007a.Fischer, E. M., Seneviratne, S. I., Vidale, P. L., Luethi, D., and Schaer,
C.:
Soil moisture – Atmosphere interactions during the 2003 European summer heat
wave, J. Climate, 20, 5081–5099, 10.1175/JCLI4288.1,
2007b.Guillod, B. P., Orlowsky, B., Miralles, D. G., Teuling, A. J., and
Seneviratne,
S. I.: Reconciling spatial and temporal soil moisture effects on afternoon
rainfall, Nature Communications, 6, 6443, 10.1038/ncomms7443,
2015.Guo, Z., Dirmeyer, P. A., Koster, R. D., Bonan, G., Chan, E., Cox, P.,
Gordon,
C. T., Kanae, S., Kowalczyk, E., Lawrence, D., Liu, P., Lu, C.-H., Malyshev,
S., McAvaney, B., McGregor, J. L., Mitchell, K., Mocko, D., Oki, T., Oleson,
K. W., Pitman, A., Sud, Y. C., Taylor, C. M., Verseghy, D., Vasic, R., Xue,
Y., and Yamada, T.: GLACE: The Global Land-Atmosphere Coupling Experiment.
Part II: Analysis, J. Hydrometeorol., 7, 611–625,
10.1175/JHM511.1, 2006.Hauser, M., Orth, R., and Seneviratne, S. I.: Role of soil moisture versus
recent climate change for the 2010 heat wave in western Russia, Geophys.
Res. Lett., 43, 2819–2826, 10.1002/2016GL068036, 2016.Hirschi, M., Seneviratne, S. I., Alexandrov, V., Boberg, F., Boroneant, C.,
Christensen, O. B., Formayer, H., Orlowsky, B., and Stepanek, P.:
Observational evidence for soil-moisture impact on hot extremes in
southeastern Europe, Nat. Geosci.e, 4, 17–21,
10.1038/NGEO1032, 2011.Hurrell, J. W., Holland, M. M., Gent, P. R., Ghan, S., Kay, J. E., Kushner,
P. J., Lamarque, J. F., Large, W. G., Lawrence, D., Lindsay, K., Lipscomb,
W. H., Long, M. C., Mahowald, N., Marsh, D. R., Neale, R. B., Rasch, P.,
Vavrus, S., Vertenstein, M., Bader, D., Collins, W. D., Hack, J. J., Kiehl,
J., and Marshall, S.: The Community Earth System Model A Framework for
Collaborative Research, B. Am. Meteorol. Soc.,
94, 1339–1360, 10.1175/BAMS-D-12-00121.1, 2013.Jaeger, E. B. and Seneviratne, S. I.: Impact of soil moisture-atmosphere
coupling on European climate extremes and trends in a regional climate
model, Clim. Dynam., 36, 1919–1939,
10.1007/s00382-010-0780-8, 2011.Koster, R. and Suarez, M.: Soil moisture memory in climate models, J.
Hydrometeorol., 2, 558–570,
10.1175/1525-7541(2001)002<0558:SMMICM>2.0.CO;2, 2001.Koster, R., Suarez, M., and Heiser, M.: Variance and predictability of
precipitation at seasonal-to-interannual timescales, J.
Hydrometeorol., 1, 26–46,
10.1175/1525-7541(2000)001<0026:VAPOPA>2.0.CO;2, 2000.Koster, R., Dirmeyer, P., Guo, Z., Bonan, G., Chan, E., Cox, P., Gordon, C.,
Kanae, S., Kowalczyk, E., Lawrence, D., Liu, P., Lu, C., Malyshev, S.,
McAvaney, B., Mitchell, K., Mocko, D., Oki, T., Oleson, K., Pitman, A., Sud,
Y., Taylor, C., Verseghy, D., Vasic, R., Xue, Y., Yamada, T., and Team, G.:
Regions of strong coupling between soil moisture and precipitation,
Science, 305, 1138–1140, 10.1126/science.1100217, 2004.Koster, R. D., Guo, Z., Dirmeyer, P. A., Bonan, G., Chan, E., Cox, P.,
Davies,
H., Gordon, C. T., Kanae, S., Kowalczyk, E., Lawrence, D., Liu, P., Lu,
C.-H., Malyshev, S., McAvaney, B., Mitchell, K., Mocko, D., Oki, T., Oleson,
K. W., Pitman, A., Sud, Y. C., Taylor, C. M., Verseghy, D., Vasic, R., Xue,
Y., and Yamada, T.: GLACE: The Global Land-Atmosphere Coupling Experiment.
Part I: Overview, J. Hydrometeorol., 7, 590–610,
10.1175/JHM510.1, 2006.Laio, F., Porporato, A., Ridolfi, L., and Rodriguez-Iturbe, I.: Plants in
water-controlled ecosystems: active role in hydrologic processes and response
to water stress – II. Probabilistic soil moisture dynamics, Adv.
Water Resour., 24, 707–723, 10.1016/S0309-1708(01)00005-7,
2001.Lawrence, D., Oleson, K., Flanner, M., Thorton, P., Swenson, S., Lawrence,
P.,
Zeng, X., Yang, Z.-L., Levis, S., Skaguchi, K., Bonan, G., and Slater, A.:
Parameterization Improvements and Functional and Structural Advances in
Version 4 of the Community Land Model, J. Adv. Model. Earth
Syst., 3, 1–27, 10.1029/2011MS000045, 2011.Lorenz, R., Davin, E. L., and Seneviratne, S. I.: Modeling land-climate
coupling in Europe: Impact of land surface representation on climate
variability and extremes, J. Geopyhs. Res.-Atmos.,
117, D20, 10.1029/2012JD017755, 2012.Lorenz, R., Argueeso, D., Donat, M. G., Pitman, A. J., van den Hurk, B.,
Berg,
A., Lawrence, D. M., Cheruy, F., Ducharne, A., Hagemann, S., Meier, A.,
Milly, P. C. D., and Seneviratne, S. I.: Influence of land-atmosphere
feedbacks on temperature and precipitation extremes in the GLACE-CMIP5
ensemble, J. Geopyhs. Res.-Atmos., 121, 607–623,
10.1002/2015JD024053, 2016.Meinshausen, M., Smith, S. J., Calvin, K., Daniel, J. S., Kainuma, M. L. T.,
Lamarque, J.-F., Matsumoto, K., Montzka, S. A., Raper, S. C. B., Riahi, K.,
Thomson, A., Velders, G. J. M., and van Vuuren, D. P. P.: The RCP greenhouse
gas concentrations and their extensions from 1765 to 2300, Climate Change,
109, 213–241, 10.1007/s10584-011-0156-z, 2011.
Oleson, K. W., Lawrence, D. M., Bonan, G. B., Flanner, M. G., Kluzek, E.,
Lawrence, P. J., Levis, S., Swenson, S. C., Thornton, P. E., Dai, A., Decker,
M., Dickinson, R., Feddema, J., Heald, C. L., Hoffman, F., Lamarque, J.-F.,
Mahowald, N., Niu, G.-Y., Qian, T., Randerson, J., Running, S., Sakaguchi,
K., Slater, A., Stöckli, R., Wang, A., Yang, Z.-L., Zeng, X., and Zeng, X.:
Technical Description of version 4.0 of the Community Land Model (CLM),
Tech. rep., National Center for Atmospheric Research, Boulder, Colorado,
2010.
Oleson, K. W., Lawrence, D. M., Bonan, G. B., Drewniak, B., Huang, M., Koven,
C. D., Levis, S., Li, F., Riley, W. J., Subin, Z. M., Swenson, S. C.,
Thornton, P. E., Bozbiyik, A., Fisher, R., Heald, C. L., Kluzek, E.,
Lamarque, J.-F., Lawrence, P. J., Leung, L. R., Lipscomb, W., Muszala, S.,
Ricciuto, D. M., Sacks, W., Sun, Y., Tang, J., and Yang, Z.-L.: Technical
Description of version 4.5 of the Community Land Model (CLM), Tech. rep.,
National Center for Atmospheric Research, Boulder, Colorado, 2013.Orth, R. and Seneviratne, S. I.: Analysis of soil moisture memory from
observations in Europe, J. Geopyhs. Res.-Atmos.,
117, D15, 10.1029/2011JD017366, 2012.Orth, R. and Seneviratne, S.: Variability of soil moisture and sea surface
temperatures similarly important for climate in the warm season, J.
Climate, 30, 2141–2162, 10.1175/JCLI-D-15-0567.1,
2017.Pitman, A.: The evolution of, and revolution in, land surface schemes
designed
for climate models, Int. J. Climatol., 23,
479–510, 10.1002/joc.893, 2003.Reale, O. and Dirmeyer, P.: Modeling the effect of land surface evaporation
variability on precipitation variability. Part I: General response, J.
Hydrometeorol., 3, 433–450,
10.1175/1525-7541(2002)003<0433:MTEOLS>2.0.CO;2, 2002.Rowell, D. P. and Jones, R. G.: Causes and uncertainty of future summer
drying
over Europe, Clim. Dynam., 27, 281–299,
10.1007/s00382-006-0125-9, 2006.Sellers, P., Dickinson, R., Randall, D., Betts, A., Hall, F., Berry, J.,
Collatz, G., Denning, A., Mooney, H., Nobre, C., Sato, N., Field, C., and
Henderson-Sellers, A.: Modeling the exchanges of energy, water, and carbon
between continents and the atmosphere, Science, 275, 502–509,
10.1126/science.275.5299.502, 1997.Seneviratne, S. I., Koster, R. D., Guo, Z., Dirmeyer, P. A., Kowalczyk, E.,
Lawrence, D., Liu, P., Lu, C.-H., Mocko, D., Oleson, K. W., and Verseghy, D.:
Soil moisture memory in AGCM simulations: Analysis of global land-atmosphere
coupling experiment (GLACE) data, J. Hydrometeorol., 7,
1090–1112, 10.1175/JHM533.1, 2006a.
Seneviratne, S. I., Luethi, D., Litschi, M., and Schaer, C.: Land-atmosphere
coupling and climate change in Europe, Nature, 443, 205–209,
10.1038/nature05095, 2006b.Seneviratne, S. I., Corti, T., Davin, E. L., Hirschi, M., Jaeger, E. B.,
Lehner, I., Orlowsky, B., and Teuling, A. J.: Investigating soil
moisture-climate interactions in a changing climate: A review, Earth-Sci.
Rev., 99, 125–161, 10.1016/j.earscirev.2010.02.004, 2010.Seneviratne, S. I., Wilhelm, M., Stanelle, T., van den Hurk, B., Hagemann,
S.,
Berg, A., Cheruy, F., Higgins, M. E., Meier, A., Brovkin, V., Claussen, M.,
Ducharne, A., Dufresne, J.-L., Findell, K. L., Ghattas, J., Lawrence, D. M.,
Malyshev, S., Rummukainen, M., and Smith, B.: Impact of soil
moisture-climate feedbacks on CMIP5 projections: First results from the
GLACE-CMIP5 experiment, Geophys. Res. Lett., 40, 5212–5217,
10.1002/grl.50956, 2013.Shukla, J. and Mintz, Y.: Influence of land-surface evapotranspiration on the
earth's climate, Science, 215, 1498–1501,
10.1126/science.215.4539.1498, 1982.van den Hurk, B., Kim, H., Krinner, G., Seneviratne, S. I., Derksen, C., Oki,
T., Douville, H., Colin, J., Ducharne, A., Cheruy, F., Viovy, N., Puma, M.
J., Wada, Y., Li, W., Jia, B., Alessandri, A., Lawrence, D. M., Weedon, G.
P., Ellis, R., Hagemann, S., Mao, J., Flanner, M. G., Zampieri, M., Materia,
S., Law, R. M., and Sheffield, J.: LS3MIP (v1.0) contribution to CMIP6: the
Land Surface, Snow and Soil moisture Model Intercomparison Project – aims,
setup and expected outcome, Geosci. Model Dev., 9, 2809–2832,
10.5194/gmd-9-2809-2016, 2016.Vautard, R., Yiou, P., D'Andrea, F., de Noblet, N., Viovy, N., Cassou, C.,
Polcher, J., Ciais, P., Kageyama, M., and Fan, Y.: Summertime European heat
and drought waves induced by wintertime Mediterranean rainfall deficit,
Geophys. Res. Lett., 34, 7, 10.1029/2006GL028001,
2007.Vogel, M., Orth, R., Cheruy, F., Hagemann, S., Lorenz, R., Hurk, B., and
Seneviratne, S.: Regional amplification of projected changes in extreme
temperatures strongly controlled by soil moisture-temperature feedbacks,
Geophys. Res. Lett., 44, 1511–1519, 10.1002/2016GL071235, 2017.Whan, K., Zscheischler, J., Orth, R., Shongwe, M., Rahimi, M., Asare, E. O.,
and Seneviratne, S. I.: Impact Of Soil Moisture On Extreme Maximum
Temperatures In Europe, Weather and Climate Extremes, 9, 57–67,
10.1016/j.wace.2015.05.001, 2015.
Wilks, D. S.: Statistical methods in the atmospheric sciences, vol. 100,
Academic press, 2011.Wilks, D. S.: The stippling shows statistically significant gridpoints
How Research Results are Routinely Overstated and Over-interpreted, and What
to Do About It, B. Am. Meteorol. Soc., 97, 2263–2273, 10.1175/bams-d-15-00267.1,
2016.